Nonlinear aeroelastic ROMs based on ML or AI algorithms can be complex and computationally
demanding to train, meaning that for practical aeroelastic applications, the conservative nature of
linearization is often favoured. Therefore, there is a requirement for novel NL aeroelastic model
reduction approaches that are accurate, simple and, most importantly, efficient to generate. Recently
our team has developed novel formulation for the identification of a compact multi-input Volterra
series, where Orthogonal Matching Pursuit is used to obtain a set of optimally sparse nonlinear multi-
input ROM coefficients from unsteady aerodynamic training data. The novel approach for the
identification of OS higher-order polynomial-based aeroelastic ROM significantly reduces the amount of
training data needed without sacrificing fidelity. The framework is exemplified using AGARD 445.6 Wing
& BSCW benchmark supercritical wing, considering forced response, flutter, and limit cycle oscillation.
In recent years, Computational Fluid Dynamics (CFD) solvers incorporating Large Eddy Simulation (LES)
turbulence models have gained significant traction. This shift is driven by the exponential growth in
computational power, enabling more accurate and detailed simulations of turbulent flows. LES models
are particularly attractive due to their ability to resolve large-scale turbulent structures while filtering the
smaller scales, providing a balance between accuracy and computational efficiency. Among the various
CFD methodologies, the Lattice Boltzmann Method (LBM) has emerged as a robust alternative for fluid
flow simulations. Unlike traditional methods that solve the Navier-Stokes equations directly, LBM
focuses on the evolution of particle distribution functions on a discrete lattice grid. This approach is
inherently compatible with parallel computing, making it highly suitable for modern computational
architectures such as Graphics Processing Units (GPUs).
June 18 2025